update: Ensure method _asdict also converts Word object into a native dict

faster_whisper/transcribe.py

@@ -5,7 +5,17 @@
 import zlib
 
 from inspect import signature
-from typing import BinaryIO, Iterable, List, NamedTuple, Optional, Tuple, Union
+from typing import (
+    Any,
+    BinaryIO,
+    Dict,
+    Iterable,
+    List,
+    NamedTuple,
+    Optional,
+    Tuple,
+    Union,
+)
 
 import ctranslate2
 import numpy as np
@@ -30,7 +40,7 @@ class Word(NamedTuple):
     probability: float
 
 
-class Segment(NamedTuple):
+class NamedTupleSegment(NamedTuple):
     id: int
     seek: int
     start: float
@@ -44,6 +54,26 @@ class Segment(NamedTuple):
     words: Optional[List[Word]]
 
 
+class Segment(NamedTupleSegment):
+    def _asdict(self) -> Dict[str, Any]:
+        words: Optional[List[Dict[str, Any]]] = None
+        if self.words:
+            words = [word._asdict() for word in self.words]
+        return {
+            "id": self.id,
+            "seek": self.seek,
+            "start": self.start,
+            "end": self.end,
+            "text": self.text,
+            "tokens": self.tokens,
+            "temperature": self.temperature,
+            "avg_logprob": self.avg_logprob,
+            "compression_ratio": self.compression_ratio,
+            "no_speech_prob": self.no_speech_prob,
+            "words": words,
+        }
+
+
 class TranscriptionOptions(NamedTuple):
     beam_size: int
     best_of: int

tests/test_transcribe.py

@@ -1,6 +1,7 @@
 import os
 
 from faster_whisper import WhisperModel, decode_audio
+from faster_whisper.transcribe import Segment, Word
 
 
 def test_supported_languages():
@@ -97,3 +98,48 @@ def test_stereo_diarization(data_dir):
     segments, _ = model.transcribe(right)
     transcription = "".join(segment.text for segment in segments).strip()
     assert transcription == "The horizon seems extremely distant."
+
+
+def test_segment_as_dict():
+    segment = Segment(
+        id=1,
+        seek=398,
+        start=0.0,
+        end=3.04,
+        text=" Hello world.",
+        tokens=[50364, 2425, 1002, 13, 50516],
+        temperature=0.9,
+        avg_logprob=-1.1086603999137878,
+        compression_ratio=0.6,
+        no_speech_prob=0.10812531411647797,
+        words=[
+            Word(start=0.0, end=1.82, word=" Hello", probability=0.8265082836151123),
+            Word(start=1.82, end=3.04, word=" world.", probability=0.31053248047828674),
+        ],
+    )
+    assert segment._asdict() == {
+        "id": 1,
+        "seek": 398,
+        "start": 0.0,
+        "end": 3.04,
+        "text": " Hello world.",
+        "tokens": [50364, 2425, 1002, 13, 50516],
+        "temperature": 0.9,
+        "avg_logprob": -1.1086603999137878,
+        "compression_ratio": 0.6,
+        "no_speech_prob": 0.10812531411647797,
+        "words": [
+            {
+                "start": 0.0,
+                "end": 1.82,
+                "word": " Hello",
+                "probability": 0.8265082836151123,
+            },
+            {
+                "start": 1.82,
+                "end": 3.04,
+                "word": " world.",
+                "probability": 0.31053248047828674,
+            },
+        ],
+    }